Data analysis To calculate determination coefficients (R2) as the explanations of wind speed, diaspore attributes and vegetation features to the proportion of diaspores dispersed more than 5 m, a permutation test was conducted with R using the ˇ°lmpˇ± function in the package lmPerm. To compare the explanations of different vegetation features, different data sets should be arranged for analysis. Data set B for comparing the explanations of vegetation coverage (10-30%) and life-form consisted of H(U) and S(U) with three levels of coverage (10, 20 and 30%) and had the same horizontal and vertical pattern but different life-form and coverage. Data set C for comparing the explanations of vegetation coverage (10-30%) and horizontal pattern consisted of H+S(U) and H+S(A) with three levels of coverage (10, 20 and 30%) and had the same life-form and vertical pattern but different horizontal pattern and coverage. Data set D for comparing the explanations of vegetation coverage (10-30%) and vertical pattern consisted of H(U), S(U) and H+S(U) with three levels of coverage (10, 20 and 30%) and had the same horizontal pattern but different vertical pattern and coverage. Data set A for calculation of the explanations of vegetation coverage (0-30%) consisted of H(U), S(U), H+S(U), H+S(A) and bare land with four levels of coverage (0, 10, 20 and 30%). To test the significant difference in proportion of diaspores dispersed more than 5 m in varying vegetation coverage, life-form composition, and horizontal and vertical pattern under each wind speed, ANOVA was conducted by rank-sum test (Kruskal-Wallis) using IBM SPSS Statistics 22.0 (IBM Corporation 1989 and 2013, USA). To test the significant interaction effects of vegetation coverage and vegetation life-form composition, horizontal pattern and vertical pattern on the proportion of diaspores dispersed more than 5 m, a permutation test was conducted with R using the ˇ°aovpˇ± function in the package lmPerm.